High Signal Podcasts Evidence ledger
Method
Browse
How the evidence works
← All source episodes

Lex Fridman Podcast / episode intelligence

#434 – Aravind Srinivas: Perplexity CEO on Future of AI, Search & the Internet

19 Jun 2024 38 published claims 2 attributable people
Open original episode ↗ Read the evidence

Speakers in the public record

Lex FridmanAravind Srinivas

Claim mix

belief 18evaluation 10uncertainty 3observation 3preference 2prediction 1recommendation 1

Evidence policy

Every row below preserves an exact excerpt. Identified speakers are linked; unresolved voices are labeled and excluded from people counts.

Claim ledger

The useful parts, with receipts.

38 published records

01 / evaluation

This significantly reduces LLM hallucinations, and makes it much easier and more reliable to use for research, and general curiosity-driven late night rabbit hole explorations that I often engage in. I highly recommend you try it out.

“This significantly reduces LLM hallucinations, and makes it much easier and more reliable to use for research, and general curiosity-driven late night rabbit hole explorations that I often engage in. I highly recommend you try it out.”
Speaker
Lex Fridman
Publisher
Lex Fridman Podcast

02 / belief

If you’ve written exams in undergrad or grad school where people allowed you to come with your notes to the exam, versus no notes allowed, I think not the same set of people end up scoring number one on both.

“If you’ve written exams in undergrad or grad school where people allowed you to come with your notes to the exam, versus no notes allowed, I think not the same set of people end up scoring number one on both.”
Publisher
Lex Fridman Podcast

03 / uncertainty

Every day I wake up in sweat thinking about how things are going to go wrong.” Because one thing you got to understand, hardware is, you got to actually, I don’t know about the 10, 20 year thing, but you actually do need to plan two years in advance because it does take time to fabricate and get the chip back and you need to have the architecture ready.

“Every day I wake up in sweat thinking about how things are going to go wrong.” Because one thing you got to understand, hardware is, you got to actually, I don’t know about the 10, 20 year thing, but you actually do need to plan two years in advance because it does take time to fabricate and get the chip back and you need to have the architecture ready.”
Publisher
Lex Fridman Podcast

05 / uncertainty

Can you have a conversation with an AI where it feels like you talked to Einstein or Feynman? Where you ask them a hard question, they’re like, I don’t know.

“Can you have a conversation with an AI where it feels like you talked to Einstein or Feynman? Where you ask them a hard question, they’re like, I don’t know.”
Publisher
Lex Fridman Podcast

06 / belief

I’m not saying it doesn’t have its flaws and there are things that are widely debated, but here I think you can just appear not to have any true emotional connection.

“I’m not saying it doesn’t have its flaws and there are things that are widely debated, but here I think you can just appear not to have any true emotional connection.”
Publisher
Lex Fridman Podcast

07 / belief

I think he went through many years where people just ridiculed or didn’t respect his work as much as they should have, and he still stuck with it.

“I think he went through many years where people just ridiculed or didn’t respect his work as much as they should have, and he still stuck with it.”
Publisher
Lex Fridman Podcast

08 / belief

My wife is insanely supportive of this journey. It’s almost like she cares equally about Perplexity as I do, uses the product as much or even more, gives me a lot of feedback and any setbacks that she’s already warning me of potential blind spots, and I think that really helps.

“My wife is insanely supportive of this journey. It’s almost like she cares equally about Perplexity as I do, uses the product as much or even more, gives me a lot of feedback and any setbacks that she’s already warning me of potential blind spots, and I think that really helps.”
Publisher
Lex Fridman Podcast

09 / belief

Now I think he has this mentality that it all has to be in-house, but it frees you from working on problems that you don’t need to be working on when you’re scaling up your startup.

“Now I think he has this mentality that it all has to be in-house, but it frees you from working on problems that you don’t need to be working on when you’re scaling up your startup.”
Publisher
Lex Fridman Podcast

10 / belief

I think another dimension is, we are just deeply curious as a species, and I think we have… Some work in AI’s, have explored this curiosity driven exploration.

“I think another dimension is, we are just deeply curious as a species, and I think we have… Some work in AI’s, have explored this curiosity driven exploration.”
Publisher
Lex Fridman Podcast

12 / uncertainty

Does that open up possibilities when you start getting to a hundred thousand tokens, a million tokens, 10 million tokens, a hundred million … I don’t know where you can go.

“Does that open up possibilities when you start getting to a hundred thousand tokens, a million tokens, 10 million tokens, a hundred million … I don’t know where you can go.”
Speaker
Lex Fridman
Publisher
Lex Fridman Podcast

13 / observation

If an AI had given him that insight when he started the company itself said, “Look, Elon, I know you’re going to work hard on Falcon, but you need to redesign it for higher payloads and this is the way to go.

“If an AI had given him that insight when he started the company itself said, “Look, Elon, I know you’re going to work hard on Falcon, but you need to redesign it for higher payloads and this is the way to go.”
Publisher
Lex Fridman Podcast

14 / belief

I think LLMs really help there. So what LLMs add is even if your initial retrieval doesn’t have a amazing set of documents, like it has really good recall but not as high a precision, LLMs can still find a needle in the haystack and traditional search cannot, because they’re all about precision and recall simultaneously.

“I think LLMs really help there. So what LLMs add is even if your initial retrieval doesn’t have a amazing set of documents, like it has really good recall but not as high a precision, LLMs can still find a needle in the haystack and traditional search cannot, because they’re all about precision and recall simultaneously.”
Publisher
Lex Fridman Podcast

15 / belief

If we can achieve that, that amount of inference compute, where it leads to a dramatically better answer as you apply more inference compute, I think that will be the beginning of real reasoning breakthroughs.

“If we can achieve that, that amount of inference compute, where it leads to a dramatically better answer as you apply more inference compute, I think that will be the beginning of real reasoning breakthroughs.”
Publisher
Lex Fridman Podcast

17 / belief

I would say in the early days of using a product, there’s a anxiety when it’s too simple because you feel like you don’t know the full set of features, you don’t know what to do.

“I would say in the early days of using a product, there’s a anxiety when it’s too simple because you feel like you don’t know the full set of features, you don’t know what to do.”
Speaker
Lex Fridman
Publisher
Lex Fridman Podcast

18 / belief

Human-computer interaction. That’s what helped them arrive with new insights to search than people who are just working on NLP, so I think that’s another thing that realized that new insights and people who are able to make new connections are likely to be a good founder too.

“Human-computer interaction. That’s what helped them arrive with new insights to search than people who are just working on NLP, so I think that’s another thing that realized that new insights and people who are able to make new connections are likely to be a good founder too.”
Publisher
Lex Fridman Podcast

19 / belief

I think he’s controversial for whatever’s happened in social media in general, but I think his positioning of Meta and himself leading from the front in AI, open sourcing, create models, not just random models, really, Llama-3-70B is a pretty good model.

“I think he’s controversial for whatever’s happened in social media in general, but I think his positioning of Meta and himself leading from the front in AI, open sourcing, create models, not just random models, really, Llama-3-70B is a pretty good model.”
Publisher
Lex Fridman Podcast

20 / belief

I think even in 2010s, one third of Google traffic, when it used to be like 3 billion queries a day, was just instant answers from- … just answers, instant answers from the Google Knowledge Graph, which is basically from the Freebase and Wikidata stuff.

“I think even in 2010s, one third of Google traffic, when it used to be like 3 billion queries a day, was just instant answers from- … just answers, instant answers from the Google Knowledge Graph, which is basically from the Freebase and Wikidata stuff.”
Publisher
Lex Fridman Podcast

22 / belief

Again, I’m kind of beating the drum here, but for me it’s all about curiosity and knowledge. And I think there are different ways to keep the light of consciousness, preserving it, and we all can go about in different paths.

“Again, I’m kind of beating the drum here, but for me it’s all about curiosity and knowledge. And I think there are different ways to keep the light of consciousness, preserving it, and we all can go about in different paths.”
Publisher
Lex Fridman Podcast

23 / belief

While I think Perplexity is trying to integrate realtime, like recent information, put priority on recent information, that’s a lot of work to integrate.

“While I think Perplexity is trying to integrate realtime, like recent information, put priority on recent information, that’s a lot of work to integrate.”
Speaker
Lex Fridman
Publisher
Lex Fridman Podcast

24 / belief

If you work from that perspective, I think you’ll give up beyond the point because it’s very hard to work towards something that was not truly important to you.

“If you work from that perspective, I think you’ll give up beyond the point because it’s very hard to work towards something that was not truly important to you.”
Publisher
Lex Fridman Podcast

27 / evaluation

A lot of startups have this problem and initially they just do things that don’t scale in the Paul Graham way, but then things start breaking more and more as you scale.

“A lot of startups have this problem and initially they just do things that don’t scale in the Paul Graham way, but then things start breaking more and more as you scale.”
Publisher
Lex Fridman Podcast

29 / evaluation

If you’re just trying to take on Google by building another [inaudible 00:15:38] search engine and with some other differentiation, which could be privacy, or no ads, or something like that, it’s not enough. It’s very hard to make a real difference in just making a better [inaudible 00:15:55] search engine than Google, because they have basically nailed this game for like 20 years.

“If you’re just trying to take on Google by building another [inaudible 00:15:38] search engine and with some other differentiation, which could be privacy, or no ads, or something like that, it’s not enough. It’s very hard to make a real difference in just making a better [inaudible 00:15:55] search engine than Google, because they have basically nailed this game for like 20 years.”
Publisher
Lex Fridman Podcast

30 / evaluation

What is the weakness of Google is that any ad unit that’s less profitable than a link, or any ad unit that kind of disincentivizes the link click is not in their interest to go aggressive on, because it takes money away from something that’s higher margins.

“What is the weakness of Google is that any ad unit that’s less profitable than a link, or any ad unit that kind of disincentivizes the link click is not in their interest to go aggressive on, because it takes money away from something that’s higher margins.”
Publisher
Lex Fridman Podcast

32 / evaluation

LLM means large language model. That LLM takes the relevant paragraphs, looks at the query, and comes up with a well-formatted answer with appropriate footnotes to every sentence it says, because it’s been instructed to do so, it’s been instructed with that one particular instruction, given a bunch of links and paragraphs, write a concise answer for the user, with the appropriate citation.

“LLM means large language model. That LLM takes the relevant paragraphs, looks at the query, and comes up with a well-formatted answer with appropriate footnotes to every sentence it says, because it’s been instructed to do so, it’s been instructed with that one particular instruction, given a bunch of links and paragraphs, write a concise answer for the user, with the appropriate citation.”
Publisher
Lex Fridman Podcast

33 / prediction

You couldn’t search over them before, but now you can because you can have a model that looks at your question, translates it to some SQL query, runs it against the database.

“You couldn’t search over them before, but now you can because you can have a model that looks at your question, translates it to some SQL query, runs it against the database.”
Publisher
Lex Fridman Podcast

34 / preference

Or like in the mobile app when you’re clicking, when you’re touching the search bar, the speed at which the keypad appears, we focus on all these details, we track all these latencies and that’s a discipline that came to us because we really admired Google.

“Or like in the mobile app when you’re clicking, when you’re touching the search bar, the speed at which the keypad appears, we focus on all these details, we track all these latencies and that’s a discipline that came to us because we really admired Google.”
Publisher
Lex Fridman Podcast

35 / evaluation

He learned the mistakes that, if you rely on others a lot for your distribution, his first company, Zip2 where he tried to build something like a Google Maps, he ended up, as in, the company ended up making deals with putting their technology on other people’s sites and losing direct relationship with the users because that’s good for your business.

“He learned the mistakes that, if you rely on others a lot for your distribution, his first company, Zip2 where he tried to build something like a Google Maps, he ended up, as in, the company ended up making deals with putting their technology on other people’s sites and losing direct relationship with the users because that’s good for your business.”
Publisher
Lex Fridman Podcast

36 / recommendation

Absolutely. And these are the kind of architectures we should explore more, where small models… And this is also why I believe open source is important, because at least it gives you a good base model to start with and try different experiments in the post-training phase to see if you can just specifically shape these models for being good reasoners.

“Absolutely. And these are the kind of architectures we should explore more, where small models… And this is also why I believe open source is important, because at least it gives you a good base model to start with and try different experiments in the post-training phase to see if you can just specifically shape these models for being good reasoners.”
Publisher
Lex Fridman Podcast

37 / preference

The Google presents this information in cool widgets, and I think that is where this is a very different problem from just building another chat bot.

“The Google presents this information in cool widgets, and I think that is where this is a very different problem from just building another chat bot.”
Publisher
Lex Fridman Podcast
Search evidence